Skip to content

feat(hpc): Session C — bgz17 dual-path integration bridge#22

Merged
AdaWorldAPI merged 1 commit into
masterfrom
claude/continue-session-0mAVa
Mar 22, 2026
Merged

feat(hpc): Session C — bgz17 dual-path integration bridge#22
AdaWorldAPI merged 1 commit into
masterfrom
claude/continue-session-0mAVa

Conversation

@AdaWorldAPI
Copy link
Copy Markdown
Owner

feat(hpc): Session C — bgz17 dual-path integration bridge

Add four new modules implementing the bgz17 Base17 encoding pipeline
directly in ndarray for interoperability without external crate deps:

  • bgz17_bridge: Base17/SpoBase17/PaletteEdge types with golden-step
    octave encoding, L1/weighted-L1 distance, XOR bind, bundle, permute.
    Matches lance-graph bgz17 crate encoding exactly.

  • palette_distance: k-means palette builder (farthest-first init),
    nearest-codebook lookup, precomputed k×k distance matrices for O(1)
    SPO distance. 128KB matrix fits L1 cache.

  • layered_distance: Container W125 palette-edge read/write, W4-W5
    truth-value read/write, TruthGate evidence filter with NARS-style
    expectation (confidence * (frequency - 0.5) + 0.5).

  • parallel_search: Dual-path HHTL + CLAM search with archetype-based
    pruning, merge-and-rerank deduplication, TruthGate filtering, and
    local fractal dimension estimation from palette distances.

All modules have comprehensive tests (880 total tests pass).

https://claude.ai/code/session_01CdqyUTUfjKZuk8YGJzv6LB

Add four new modules implementing the bgz17 Base17 encoding pipeline
directly in ndarray for interoperability without external crate deps:

- bgz17_bridge: Base17/SpoBase17/PaletteEdge types with golden-step
  octave encoding, L1/weighted-L1 distance, XOR bind, bundle, permute.
  Matches lance-graph bgz17 crate encoding exactly.

- palette_distance: k-means palette builder (farthest-first init),
  nearest-codebook lookup, precomputed k×k distance matrices for O(1)
  SPO distance. 128KB matrix fits L1 cache.

- layered_distance: Container W125 palette-edge read/write, W4-W5
  truth-value read/write, TruthGate evidence filter with NARS-style
  expectation (confidence * (frequency - 0.5) + 0.5).

- parallel_search: Dual-path HHTL + CLAM search with archetype-based
  pruning, merge-and-rerank deduplication, TruthGate filtering, and
  local fractal dimension estimation from palette distances.

All modules have comprehensive tests (880 total tests pass).

https://claude.ai/code/session_01CdqyUTUfjKZuk8YGJzv6LB
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants